Build `tool_use_with_validation(tool_def, args)` that mirrors the safety check senior engineers wrap around every tool call: 1. Read `tool_def["input_schema"]` (JSON Schema dict with `properties` and `required`). 2. If any `required` key is missing from `args` → return `{"valid": False, "error": "missing:<key>"}`. 3. For each arg present, if the property declares a `type`, check it matches (`"string"`→str, `"integer"`→int, `"number"`→float, `"boolean"`→bool). On mismatch → `{"valid": False, "error": "type:<key>"}`. 4. Otherwise → `{"valid": True, "result": "called <name> with <sorted args>"}`. The harness runs 3 scenarios: a clean call, a missing required field, and a type mismatch.